跳到正文
arXiv cs.AI· Zaer Bin Zaqy, Brinnae Bent·· 23 小时前AI 评分22

通过解码器进行归因:面向结构化输出的预测空间解释方法

Attribution Through the Decoder: Prediction-Space Explanations for Structured Outputs

AI 导读

针对角度、视线方向等结构化输出,研究提出通过解码器微分进行归因,而非直接对原始坐标做归因,以避免把仅影响缩放、不影响方向的输入误判为重要。在 MPIIGaze 验证集上,tangent Grad-CAM 在删除 10% 像素时预测位移比基线高 44%;在 PASCAL3D+ 的 6D 旋转解码器上,完整微分方法相比原始坐标求和平均删除位移提升 14.1 度。

正文 · 原文

View PDF HTML (experimental)

Abstract:Networks that predict angles, gaze directions, and rotations do not output these quantities directly: they emit raw coordinates that a fixed decoder, such as normalization, maps to the final prediction. Attribution methods are usually applied to the raw coordinates, so they can credit inputs for changes the decoder discards, such as rescaling a vector when only its direction matters. The resulting explanation can highlight inputs that have no effect on the prediction. We instead attribute through the decoder differential, which for normalization decoders reduces to a scaled tangent projection. We characterize invariance to output rescaling, show that every fixed smooth scalar target on a sphere has points of zero tangential sensitivity, and show that an endpoint-projected variant of Integrated Gradients can still attribute to scale-only inputs, whereas direct angular Integrated Gradients cannot. On synthetic tasks with analytically known relevance, ambient attribution assigns up to 72% of its mass to an irrelevant region. On MPIIGaze validation data, tangent Grad-CAM produces 44% greater prediction displacement under 10% pixel deletion. On PASCAL3D+ with a 6D rotation decoder, the full differential improves mean deletion displacement by 14.1 degrees over the ambient coordinate sum and by 13.8 degrees over a decoder-invariant scalar. Results for a two-polarity tangent LayerCAM variant are mixed: it exceeds ambient-sum LayerCAM on circular regression but no difference is detected on quaternion pose. These results separate two failure sources, radial contamination and scalarization loss. The approach uses existing attribution methods with decoder-aware targets, although some variants also require changes to attribution aggregation. An invariant attribution can still miss inputs that drive the prediction, so invariance and deletion performance should each be checked.
Comments: Under review
Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
Cite as: arXiv:2611.23161 [cs.CV]
  (or arXiv:2611.23161v1 [cs.CV] for this version)
  https://doi.org/10.48550/arXiv.2611.23161

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Brinnae Bent [view email]
[v1] Tue, 17 Nov 2026 18:13:50 UTC (533 KB)

来源:arXiv cs.AI · arxiv.org